采用广义可加模型分析预测上海市流感样病例发病情况  

Prediction of influenza-like illness in Shanghai based on the generalized additive method

作　　者：陈健[1,3] 张磊[2] 陆帅[2] 姜晨彦[1] 胡家瑜[1] 姜庆五[3] 吴凡[1] 

机构地区：[1]上海市疾病预防控制中心,200336 [2]复旦大学数学科学学院 [3]复旦大学公共卫生学院公共卫生安全教育部重点实验室

出　　处：《中华流行病学杂志》2013年第4期404-408,共5页Chinese Journal of Epidemiology

基　　金：上海市卫生局课题（2010188）;上海市科委生物医药重大专项（09DZ1906600）;国家自然科学青年科学基金（11101093）;上海市浦江人才计划（11PJ1400800）;2010年上海领军人才“地方队”培养计划

摘　　要：探讨广义可加模型（GAM）在上海市流感样病例发病趋势分析和预测中应用，并通过已知的气象条件预测可能的流感样病例数量。研究中利用2006-2010年上海市每周气象数据以及流感样病例监测数据，按照GAM理论，建立气象数据与流感样病例问基于非线性回归的数学模型。通过初步的数据分析，构造多个候选模型，并通过AIC（Akaike information criterion）指标选取合适的模型进行数据分析及预测。基于周平均气温及周平均日温差（周相对湿度）的模型较好拟合了原始数据，且模型简洁、明确。对于原始数据的拟合残差及部分拟合残差基本符合上海市流感样病例发病的实际情况，并具有一定的预测能力。结果表明GAM能够较好拟合上海市流感样病例发病与气象因素的变化趋势，准确预测流感样病例的发病情况，适合应用于气象因素依赖的疾病发病预测和分析。The aim of the current research topic was to test the generalized additive method （GAM） , using data from the analysis and prediction on influenza-like illness （ILI） in Shanghai. Through collecting the meteorological data as well as the ILI from 2006 to 2010, we established several nonlinear regression candidate models based on the GAM. These models considered factors as： the nonlinear dependence on the meteorological data, i.e. weekly average temperature and weekly average （maximum） temperature differences and the ILI. The AIC （Akaike information criterion） involved two simplified models which were implemented for further analysis and prediction. Finally, numerical examples showed that the proposed models could shed light on the connection between the meteorological data and the ILl. GAM could be used to fit the frequencies of ILI and meteorological factors in Shanghai. The proposed models were able to accurately analyze the onset of ILl, implying that GAM might be suitable for the prediction and analysis of those meteorological correlative diseases.

关 键 词：广义可加模型 流感样病例 预测 

分 类 号：R181[医药卫生—流行病学] R51[医药卫生—公共卫生与预防医学]